Re: PCA analysis of rep grids

Bob Green (
Sat, 8 Mar 1997 20:19:02 +1000


Regarding your comments,

>The real problem is that there is no 'true' representation of a grid. What
>your queries raise is the extent of our lack of knowledge about how the
>various methods of analysis impose their own artefactual structure on the
>solution (ph.d. anyone?)

This is an issue which has been of interest to me (though not to the extent
of doing a ph.d). Is this a one way issue or can the data significantly
affect the analysis method (a two way interaction if you like)? Out of
interest and probably scepticism I analysed four grids using a variety of
different PCA programs and analysis methods and found that for some grids
there was a high level of agreement between programs whereas for other grids
agreement was much poorer. James Grice (who may be on this list) had
suggested that amongst other factors (e.g rotation and iteration criteria),
the variance in the data may account for difference in results across grids.
I had also wondered whether types of construing (e.g., loose versus tight)
might be 'better' represented by different methods or programs.

If the method or program does make a significant difference in the results
which are obtained (and this should also apply to non-grid analyses) greater
caution should be used in conclusions which are drawn and strategies should
be used to validate results. Such a conclusion also suggests that
alternative constructivism could fruitfully be applied to the analysis and
interpretation of much research. As you note the results of a grid analysis
are a representation rather than 'objective' depiction of construing.

Regarding the demise of G-Pack, do you see standard analysis packages as the
preferred alternative for the future or do you see a role for dedicated grid
analysis packages?


Bob Green